We have now released the test sets for all tasks and also some additional resources: http://www.statmt.org/wmt18/quality-estimation-task.html
We draw your attention to task 4, which is new: taking neural machine translations of product titles and descriptions, words are annotated for errors with severity levels, and these annotations are used to put together a 'document'-level score for the set of title/description sentences. Annotations on both word and document levels are made available and the task is to predict the latter.
The deadline for submission is June 15th.
On 19 February 2018 at 18:13, Lucia Specia <lspecia at gmail.com> wrote:
> Dear all,
> The training data for the 7th edition of the shared task on
> automatic methods for estimating the quality of machine translation output
> is out!
> The main novelties are:
> 1) neural MT output
> 2) prediction of information missing in translations
> 3) human annotations for words/phrases and documents
> 4) prediction of various post-editing effort indicators, including
> post-editing time and keystrokes
> Check the website for more details: http://www.statmt.org/wmt18/
-- Lucia www.dcs.shef.ac.uk/~lucia/ -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/html Size: 3240 bytes Desc: not available URL: <https://mailman.uib.no/public/corpora/attachments/20180523/7aceda93/attachment.txt>